Unlocking Advanced Processing Capabilities with OpenTelemetry
Talkat Observability Day North America 2024 · on Tuesday, 2024-11-12· at 14:40· Salt Palace | Level 2 | 255 E·
summary
OpenTelemetry configuration examples show concrete wins: use the filter processor and OTL transforms to drop debug logs and normalize high‑cardinality attributes, convert spans into metrics with the span to metrics processor, and route noncritical data to cheap object storage. Tail sampling cuts outgoing volume to ~20% in examples but requires keeping traces in memory and a two‑tier routing design; Skyscanner reports ~2.7M spans/sec and ~100k traces/sec and runs tail sampling at ~4–5%. The failover connector enables health‑based exporter routing for HA.
AI-written summary · about
chapters (10)
- 0:00OpenTelemetry processing intro
- 1:26Why processing matters
- 4:18Collector pipelines explained
- 5:20Filtering telemetry best practices
- 9:53Sampling for cost reduction
- 15:07Routing telemetry backends
- 17:54High availability & recovery
- 18:57Transforming telemetry data
- 21:27SkyScanner cost optimization
- 23:53Q&A
related talks
22mShift Into an Observability Mindset with OpenTelemetry
33mReal-World Sampling – Lessons Learned...
46mThe State of OpenTelemetry
20mObservability and Open Telemetry
52mHow to effectively spy on your systems
28mControlling data overhead from the OpenTelmetry Collector
22mOpen Telemetry for API Monitoring
36mObservability For You and Me with OpenTelemetry